This paper explores school effects in Japan, using a Multilevel Model Analysis. "School Effects" consider how schools can affect children's learning outputs. This is one of the important subjects in the sociology of education. Since the Coleman Reports of the 1960s, there has been much research investigating school effects all over the world. This is termed "School Effectiveness Research" (SER). Recently, SER researchers are planning comparative studies within countries, and this research has considerable attraction. Unfortunately, very few researchers in Japan knew about SER until 2000. Pioneers in SER in Japan are Nabeshima and Shimizu. They investigated school effects using the concept of "Effective Schools" conceived by Edmonds, one of the most famous early SER researchers. Japanese SER researchers argue that "Effective Schools" are characterized by "teacher cooperation," "leadership," and "good classroom climates." Finally, they conclude, "If all schools had these characteristics, social disadvantage would be less than at present." However, the concept of "Effective Schools" in Japan is problematic, and thus we need to reanalyze carefully the conclusions of pioneer SER research in Japan. In this paper, I firstly point out the lack of the concept of "Effective Schools" and propose a Multilevel Model Analysis (Hierarchical Linear Modeling) to analyze school effects. Secondly, I investigate school effects in Japan using the Multilevel Model Analysis, and analyze academic achievement tests at 5th grade conducted in Z City in the district of Kansai. The sample size is 3,366 children from 43 schools and explanatory variables are achievement score at 4th grade, children's gender, and children's cultural social class. The findings are as follows. Firstly, almost all the variance in academic achievements at 5th grade exists within schools. There is small variance between schools. Variance between schools is less than 5 percent. Secondly, the most important explanatory variable affecting achievement at 5th grade is achievement at 4th grade. The most explanatory variables following this are children's gender and cultural social class. Thirdly, although the variance between schools is very small, there is a score gap of 5 to 6 points between "the most effective school" and "the most ineffective school." Thus, there are very small score gaps between schools, but this does not mean that schools have no effect on the alleviation of social disadvantage caused by gender or social class. We draw the following conclusions from this analysis. First, junior high schools in Japan have very similar learning environments, and so there are only small score gaps between schools. Secondly, the prior achievement score has a very strong effect on achievement score at the 5th grade level. Investigating school outcomes without taking prior achievement scores into account is therefore very problematic. In Japan, few argue the importance of prior achievement tests, but this is a critical point for the analysis of achievement scores. Thirdly, it is important that researchers of SER in Japan understand that there is only small variance between Japanese schools and argue seriously about "what schools can and cannot do." [ABSTRACT FROM AUTHOR]